Today, the local approach to fracture is widely applied to simulate the failure of specimens. For ductile damage processes the Gurson-Tvergaard-Needleman model is the quasi-standard. In the last time non-local extensions allowed a mesh-size independent simulation of crack growth. However, most publications dealing with this subject focus upon the convergence regarding global quantities such as the load-displacement relation. Minor attention is paid to the fields directly at the crack tip. Correspondingly, the interrelationship between the intrinsic length of the model and relevant microscopic damage processes at the crack tip is only partly established until now. In the present study the crack propagation is simulated for an implicitly gradient enriched GTN-model within a boundary layer in order to overcome influences of the specimen geometry. The different stages of damage evolution are resolved by a fine mesh.
ModelIn order to describe the ductile crack initiation and propagation a non-local extension of the widely used GURSON [1] model with the modifications of TVERGAARD and NEEDLEMAN (GTN,[2,3]) is employed. For this purpose, the volumetric plastic strain rate ε vol in the evolution equation of the local model for the void volume fraction f is replaced by its non-local counterpart ε nl using an implicit gradient enrichment [4]:Thereby, an intrinsic length scale l nl is introduced via the prefactor of the LAPLACE-operator ∆Ô Õ. The formulation can be interpreted as a spatial average over the local volumetric plastic strain ε vol around the actual point. The crack propagation behavior of the model is investigated under small-scale yielding conditions by means of a boundary layer approach. This has the advantage that a possible influence of the specimen geometry can be excluded. The hardening is described by a power law (exponent N ). In the FE-model elements with quadratic shape functions for the displacements and linear ones for the non-local strains are used, for details see [4]. Dynamic simulations are performed with implicit time integration under quasi-static loading [5].In the initial loading stages the plastic deformations occur within a region much smaller than the intrinsic length scale l nl . Hence, at this stage no void growth but pure crack tip blunting is predicted by the model which is physically reasonable. In order to handle the strain singularity at the crack tip coinciding with blunting of an ideal sharp crack, an initial radius is used at the crack tip, see fig. 1. This procedure allows to use elements with edge lengths h lig considerably smaller than the value of h lig Ô0.5 . . . 2.5Õ l nl used in the literature [6,7] where numerical problems with distorted elements were reported for finer meshes [7]. For comparison, simulations without rounded crack tip were performed as well. Fig. 1 Details of the finite element mesh at the crack tip with initial radius
l nl
ResultsFirst, simulations with a common set of parameters (f 0 0.01, f c 0.15, f f 0.25, N 0.1, E 250 σ y , ν 0.3) and several spat...